deepseek-r1-alpaca-nemotron-8b

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.8.0.dev0

base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
lora_model_dir: jrestro/deepseek-r1-nemotron-finetuned
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
tokenizer_path: deepseek-ai/DeepSeek-R1-Distill-Llama-8B

load_in_8bit: true
adapter: lora
lora_r: 16
lora_alpha: 32
lora_dropout: 0.1
lora_modules_to_save:
  - embed_tokens
  - lm_head

output_dir: ./outputs/nemotron_a40_turbo

datasets:
  - path: tatsu-lab/alpaca
    type: alpaca
    split: train
  - path: jrestro/nemotron-alpaca-converted
    type: alpaca
    split: train

val_set_size: 0.05
sequence_len: 256
micro_batch_size: 6
gradient_accumulation_steps: 1
learning_rate: 0.0003
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
num_epochs: 1
warmup_steps: 100
bf16: true
gradient_checkpointing: true
flash_attention: true
evals_per_epoch: 1
saves_per_epoch: 2
logging_steps: 10
weight_decay: 0.0

ddp: true
ddp_timeout: 7200
dataloader_num_workers: 8
dataloader_pin_memory: true

</details><br>

outputs/nemotron_a40_turbo

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B on the tatsu-lab/alpaca and the jrestro/nemotron-alpaca-converted datasets. It achieves the following results on the evaluation set:

• Loss: 1.5442

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 6
• eval_batch_size: 6
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 12
• total_eval_batch_size: 12
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 1.0

Training results

Framework versions

• PEFT 0.14.0
• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0